Four primary segments and three secondary segments. Plus explicit non-audience.
The four primary segments differ in scale and resources, but all face the same underlying problem: software projects accumulate methodology debt fast when AI agents are in the contributor mix. The specific frustrations vary by segment — collisions for small teams, drift for solos, tech debt for indies, integration ambiguity for leaders — but the root is the same: no established methodology was built natively for human + AI collaboration. The four segments below are different lenses on that one problem.
Profile: one person + one or two AI coding tools (Claude Code, Cursor, Codex, Aider, Continue), working on a personal or freelance project. Probably solo SaaS, indie game, side project, or research code.
What they’re trying to accomplish: ship a working product. Move fast. Use AI to do what would otherwise take far longer to type by hand.
Current frustrations:
What they’d pay (in attention) to fix: anything that lets them keep velocity without losing the thread. They’ll read a 12-doc methodology if it pays off; they won’t read SAFe.
Why this methodology fits: every protocol works for one person, lock TTL is generous (single contributor rarely collides with themselves), CLAUDE.md / AGENTS.md gives the AI standing context, the four planning layers prevent drift, the DoD prevents shipping half-done.
Profile: small startup engineering team, agency project team, research group, or open-source maintainers. Humans and AI agents working in the same codebase on overlapping concerns.
What they’re trying to accomplish: coordinate work across humans and AI without enterprise process overhead. Ship features. Maintain quality. Avoid the “who broke main?” question.
Current frustrations:
What they’d pay (in attention) to fix: lightweight coordination they don’t have to manually orchestrate. Locks that work for both humans and AI. A clear “who owns which decision” reference.
Why this methodology fits: file-based locks with TTL (humans and agents use the same protocol), decision-ownership matrix (concrete who-decides-what), HUMAN_NEEDED.md (blocked work surfaces visibly), DoD coupled to item frontmatter (can’t fake done).
Profile: building a product mostly with AI doing the implementation work. Limited time, limited resources, no full-time engineer.
What they’re trying to accomplish: ship something real. Validate the product. Avoid building the wrong thing fast.
Current frustrations:
What they’d pay (in attention) to fix: anything that makes the project less fragile. They especially value the working-principles (“AI agents stay on task”) and the four planning layers (“today’s commit traces back to a strategy phase”).
Why this methodology fits: the methodology is designed for projects where AI does most of the typing. The supervisory-layer framing in 11_human_roles.md is exactly the founder’s job. The Step 0 brief is what most indies skip — and what most indies later regret skipping.
Profile: team lead, eng manager, or VPE at a small-to-mid company. Existing team with established process (Scrum-ish, Kanban-ish, GitHub Issues, some kind of sprint cadence). Adding AI agents — Copilot, Cursor, Claude Code — into the workflow.
What they’re trying to accomplish: integrate AI agents without disrupting what works, without inventing the wheel, without committing to a vendor lock-in. Answer the team’s “how should we work with AI now?” question with something defensible.
Current frustrations:
What they’d pay (in attention) to fix: a defensible overlay on top of existing process that names the new failure modes (cheating agent, yes-man, stranger in own code, tribal-knowledge loss) and provides counters.
Why this methodology fits: explicit “coexists with Scrum / Kanban / XP” framing in the README; the four anti-patterns in 11_human_roles.md are exactly the failure modes leaders are worried about; the decision-ownership matrix gives a defensible answer to “what should AI decide vs what should humans decide?”
Vendors and open-source maintainers of AI coding tools (Anthropic, OpenAI, Cursor, Aider, Continue, Antigravity). Read the methodology to see what their tools should support — instruction files, plan mode, cross-session memory, etc. The methodology is a soft spec for what good AI-coding tooling enables.
Bootcamp instructors, CS educators, internal training leads. Use the methodology as curriculum reference for “how to use AI tools responsibly in real projects.” The methodology’s voice (direct, opinionated, principle-based) translates well to teaching.
Academics working on human-AI collaboration in software engineering. Use the methodology as case-study material (a real, in-use methodology with versioned history). The CHANGELOG is the kind of artifact a research paper can cite.
The 1-year horizon audience is the early adopter subset — people who actively seek methodology guidance for AI-coding. The 3-year horizon audience grows to include the mainstream subset — people who notice the named methodology as the field stabilizes.
The 3-year horizon also includes fork maintainers for niche variants (regulated industry, research teams, etc.) who didn’t exist in the 1-year window because there hadn’t been time for forks to mature.